diff --git a/examples/quantize/quantize.cpp b/examples/quantize/quantize.cpp
index 7c991c8d5..115d8fb1b 100644
--- a/examples/quantize/quantize.cpp
+++ b/examples/quantize/quantize.cpp
@@ -52,6 +52,13 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
+    // needed to initialize f16 tables
+    {
+        struct ggml_init_params params = { 0, NULL, false };
+        struct ggml_context * ctx = ggml_init(params);
+        ggml_free(ctx);
+    }
+
     // parse command line arguments
     const std::string fname_inp = argv[1];
     std::string fname_out;
diff --git a/llama.cpp b/llama.cpp
index 7b599c173..5b154d50c 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -2188,6 +2188,11 @@ struct llama_context * llama_init_from_file(
     return ctx;
 }
 
+void llama_init_ggml(struct ggml_init_params params) {
+    struct ggml_context * ctx = ggml_init(params);
+    ggml_free(ctx);
+}
+
 void llama_free(struct llama_context * ctx) {
     delete ctx;
 }
@@ -2198,12 +2203,6 @@ int llama_model_quantize(
   enum llama_ftype   ftype,
         int          nthread) {
     try {
-        // needed to initialize f16 tables
-        {
-            struct ggml_init_params params = { 0, NULL, false };
-            struct ggml_context * ctx = ggml_init(params);
-            ggml_free(ctx);
-        }
         llama_model_quantize_internal(fname_inp, fname_out, ftype, nthread);
         return 0;
     } catch (const std::string & err) {
diff --git a/llama.h b/llama.h
index f955fa23d..db98d05b7 100644
--- a/llama.h
+++ b/llama.h
@@ -97,6 +97,10 @@ extern "C" {
                              const char * path_model,
             struct llama_context_params   params);
 
+    // Init the ggml context (it won't return a context ptr because it will free
+    // the ctx after initialing it).
+    LLAMA_API void llama_init_ggml(struct ggml_init_params params);
+
     // Frees all allocated memory
     LLAMA_API void llama_free(struct llama_context * ctx);